The problem – preventing hospital-acquired venous thromboembolism
Venous thromboembolism (VTE) remains a major worldwide health burden, with an annual incidence of between 0.75 and 2.69 per 1000 individuals in the population.1 At least half of all VTE occurs during hospitalisation, or up to 90 days following discharge; such cases are described as hospital-associated VTE or hospital-acquired thrombosis (HAT).2 Many HAT cases are potentially preventable, through the use of thromboprophylaxis during the initial hospital stay and patient education.3–5 Despite the widespread availability of evidence-based prevention strategies, HAT has been described as the number one patient safety issue for individuals requiring hospital admission, worldwide.6 Previous evidence suggests that HAT accounts for more deaths and disability than nosocomial pneumonia, catheter-related bloodstream infections or adverse drug events in low- and middle-income countries.7
Both pharmacological and mechanical thromboprophylaxis can reduce the risk of VTE. However, pharmacological thromboprophylaxis potentially increases the risk of bleeding and exposes patients to iatrogenic adverse events. Therefore, the decision to provide thromboprophylaxis involves consideration of the risks of VTE and bleeding, along with the costs of providing thromboprophylaxis and treating the consequences of VTE and bleeding.
VTE risk assessment models (RAMs) are used to estimate the risk of VTE and hence the benefit of providing thromboprophylaxis. Using a VTE RAM to target those with higher risk should improve cost-effectiveness. However, this cannot be assumed. If thromboprophylaxis is very effective, then it may be cost-effective to treat everyone rather than treating only those at higher risk. If it is not very effective, then it may not be cost-effective to treat anyone.
The cost-effectiveness of VTE risk assessment tools: uncertainty and evidence deficit
Many VTE RAMs have been developed and some have been validated by estimating the prognostic accuracy for VTE. Systematic reviews have examined VTE RAMs for acutely ill medical patients (11 studies), hospitalised non-surgical patients (11 studies) and all hospitalised populations (22 studies).8–10 The reviews identified many RAMs but with limited evaluation. The primary studies had important methodological limitations and were generally judged as low quality. There was substantial variation between the primary studies in terms of populations, methods and outcomes used, which precluded meta-analysis and limited comparison between RAMs. Overall, the studies suggested that the identified RAMs had modest prognostic value with most reporting C-statistics around 0.6–0.7. Sensitivity and specificity depended upon the threshold used but high sensitivity could only be achieved by substantial loss of specificity. An ideal RAM could not be recommended by any of the reviews. However, further primary research has been published since these systematic reviews were undertaken, so there may be some benefit from updating these reviews.
To date, limited research has explored the trade-off between the risks of VTE and thromboprophylaxis. This trade-off is essential in determining whether a RAM will be cost-effective and, if so, the threshold of risk or balance of sensitivity and specificity that should be used in decision-making. Le et al. used decision-analytic modelling to estimate a risk threshold for prophylaxis in hospitalised medical patients.11 Undertaken from a US health system perspective with a willingness to pay threshold of $100,000 per quality-adjusted life-year (QALY) gained, the analysis showed that pharmacological thromboprophylaxis was cost-effective for an average medical patient with a VTE risk exceeding 1.0%. Decision-analytic modelling undertaken from a NHS perspective using thresholds advised by the National Institute for Health and Care Excellence (NICE) could inform NHS use of VTE prophylaxis.
There has been very little primary research into VTE RAMS in the NHS. Ideally, the preferred approach in this situation would be to derive and validate VTE RAMs in a typical setting. However, the established effectiveness of VTE prophylaxis in hospital inpatients presents a substantial challenge for studies estimating the accuracy of VTE RAMs. Given our current understanding, it would be unethical to withhold prophylaxis from patients who are at risk of VTE. The fundamental aim of risk assessment is to identify patients where VTE could be prevented by thromboprophylaxis. Over 70% of medical patients in the UK receive thromboprophylaxis when the Department of Health (now Department of Health and Social Care) risk assessment tool is used (NICE 2018).8 In this situation, around half of all VTE will have been prevented and any prognostic model derived from observational data will largely be predicting VTE that would not be prevented with thromboprophylaxis.4,12 Any prognostic model derived in this setting would, therefore, be based on factors that predict non-preventable VTE while underestimating (or missing) those that predict preventable VTE. A RAM that predicted non-preventable VTE while failing to predict preventable VTE would be of no clinical benefit and could potentially increase harm.
These uncertainties are highlighted in recent NICE guidance within a UK setting, which now advocates use of any RAM published by a national UK body, professional network or peer-reviewed journal.8 Australasian guidance is similar.13 However, North American guidelines offer variable messaging on risk assessment. The American Society of Haematology guidelines avoid any recommendations on RAM use, but advise that clinicians ‘integrate VTE and bleeding risk assessments into clinical decision-making processes’.14 Alternatively, the American College of Chest Physician guidelines dichotomise non-surgical patients into low and high risk within their recommendations, basing assessment on a specific RAM (the Padua prediction score).15,16 All guidelines highlight the limitations within the evidence base for RAMs. Such disparity in recommendations invites variation in practice and ongoing uncertainty for clinicians and patients.